FinOps for Apache Kafka®

See what your Kafka actually costs.
Then cut 20–40% of it.

Brod connects read-only, attributes real monthly cost to every topic and team, and turns waste into euro-quantified pull requests your engineers review. It cannot touch your cluster — by architecture.

Start monitoring →

Free, MIT-licensed CLI · read the install script first — we would too.

brod audit — prod-eu

$ brod audit --cluster prod-eu
Scanning 214 topics, 38 consumer groups… done in 4.2s (read-only)

          1
          events.clickstream.raw — retention 30d, max consumer lag 2h → reduce to 7d
          €1,240/mo est.
          2
          orders.v1.deadletter — 0 consumers, 412 GB retained → archive & delete
          €610/mo est.
          3
          telemetry.ingest.v2 — replication 3 → 2 on a non-critical topic
          €480/mo est.
        
3 of 9 findings shown · total identified: €4,180/mo
→ continuous monitoring & PR-based fixes: https://brod.sh

Read-only, by design

Write-capable credentials are rejected at onboarding. Every fix ships as a config diff, Terraform change, or GitOps PR — reviewed and merged by your engineers. Brod cannot modify a cluster.

Metadata only

Topic configs, sizes, offsets, consumer-group states. Never message payloads — there is no code path that fetches one, and the DPA says so. Security review in minutes, not months.

Euros, not metrics

Every finding carries a € figure, the exact change, and a labeled basis — estimate or measured. Savings are booked only after a fix is measured, never when it's predicted.

Connect. See the money. Merge the fix.

From read-only credentials to a merged saving in under a week.

STEP 1

Connect read-only

Self-hosted, AWS MSK, or Confluent Cloud. Deploy the collector — a 5 MB static binary, FROM scratch image — inside your network. Metadata snapshots leave over HTTPS; nothing else does.

~5 minutes

STEP 2

See cost per topic & team

Your real monthly bill, attributed to topics, teams, and owners. A rules engine flags over-retention, dead topics, excess replication, and more — each one ranked in euros per month.

first findings in ~1 hour

STEP 3

Merge the pull request

Each fix arrives as a reviewable PR with the diff and its evidence attached. After it lands, Brod measures the actual delta — and only then books the saving to your ledger.

savings booked when measured

Anatomy of a finding

A number your CFO believes. A diff your engineers approve.

No dashboards to interpret. Every finding is the euro figure, the evidence behind it, and the exact configuration change — packaged as a pull request. Nothing happens until someone on your team merges it.

R3 · OVER-RETENTION ESTIMATE

events.clickstream.raw

Retained 30 days; no consumer has read past 2 hours of lag in 14 days.

- retention.ms = 2592000000  # 30 days
+ retention.ms = 604800000   # 7 days

€1,240/mo

basis: 9.6 TB reclaimed × €0.129/GB-mo

Open pull request →

Pays for itself in the first finding.

Credit card, not procurement. EU-hosted, EU-invoiced.

CLI

Free

MIT-licensed, forever

—One-shot audit, euro-ranked
—Runs entirely in your terminal
—Same rules engine as the SaaS

curl -fsSL brod.sh | sh

MOST TEAMS

TEAM

€149/mo

up to 3 clusters · 14-day free trial

—Continuous monitoring & alerts
—Cost per topic, team & owner
—PR-based fixes with evidence
—Measured-savings ledger

Start 14-day trial

SCALE

€499/mo

up to 10 clusters

—Everything in Team
—Chargeback exports for finance
—Slack digests & API access
—Priority support

Start 14-day trial

Hosted in the EU (Hetzner, Falkenstein) · metadata only · DPA & one-click deletion included

Why “Brod”

Max Brod was Kafka’s closest friend and literary executor — he looked after everything Kafka left behind, and he knew what it was worth.

That’s the job. Brod looks after what your Kafka leaves behind — every topic, every byte, every euro — and never presumes to change a word of it.

See what your Kafka actually costs.Then cut 20–40% of it.